Font Size: a A A

Tongue Fissure Recognition

Posted on:2021-03-18Degree:MasterType:Thesis
Country:ChinaCandidate:C WanFull Text:PDF
GTID:2404330605452540Subject:Mechanical engineering
Abstract/Summary:PDF Full Text Request
Traditional Chinese medicine(TCM)is the crystallization of the wisdom and medical practice of the Chinese people,and its treatment theory and method are widely valued by people.Tongue diagnosis is an important means of TCM diagnosis,and fissure tongue is an important tongue image in tongue diagnosis.The study of fissured tongue and its characteristics are of great significance for early diagnosis of disease and health care.It is a trend to use modern technology in traditional tongue diagnosis to make tongue diagnosis more objective,standardized and intelligent.The project studied the objectification of the tongue fissure,which mainly includes:tongue image segmentation,fissured tongue recognition,fissure extraction,fissure feature extraction and recognition,and the improvement of the tongue image acquisition device.Proposed a tongue image segment method based on SegNet,after pre-processing,SegNet model were constructed to obtain the tongue body,and mathematical morphology was used to optimize the results.Proposed a fissure extraction method based on the Otsu and bot-hat,obtained fissure area by the Otsu and extracted fissure by bot-hat,deleted fake fissures by postprocessing.Extracted shape and color features of fissure,and the recognition of fissure shapes and depth were realized by using these features and Decision Tree.According to the fissure features,a quantitative analysis method for the severity of fissure tongue was proposed.The camera fixing device and light source position of the original tongue image acquisition device are redesigned and tongue images database are developed.The results show that the segmentation method based on SegNet can effectively segment tongue image with high accuracy of 92.62%.The fissured tongue was well recognized with 98.2%accuracy.The fissure extraction method was effective and meet the analysis requirements with recall of 81.4%and precision of 86.1%.And the shapes and depth of fissure were well recognized by with accuracy of 90%.According to the calculation method of fissure severity,the values of slight and serious fissure tongue are 12.3 and 36.7 respectively,which can reflect the severity of fissure tongue in some extend.The improved tongue image acquisition device has more complete functions of software and hardware.
Keywords/Search Tags:tongue image segmentation, fissured tongue recognition, fissure extraction, fissure recognition
PDF Full Text Request
Related items